[ih] OSPF convergence

[J. Noel Chiappa]

    > From: "Li, Ke Qin (Peter)" <keqinli at lucent.com>

    > I want to know: When, how, and by whom was the convergency of the OSPF
    > protocol proved.

I'm not sure anyone has proved it for OSPF in particular.

As you probably know, OSPF is an example of a broad class of algorithms
called "link state", originally invented by John McQuillan (then of BBN), and
it's possible that someone (either John, in his PhD thesis [which used to be
available from the NTIS], or perhaps some other LS effort, like IS-IS) has
proved that it converges.


I'm not sure that anyone has bothered though, because proving converge is
much less interesting (to most routing experts) than understanding the
stabilization response curve (i.e. with %-age of cases along one axis, and
stabilization time along another). I think that's probably too complex to do
as a closed-form thing (and in any case it can be dependent on the actualy
topology).

So there has been a lot of simulation to compare LS algorithms with other
kinds of routing algorithm - which had the side-effect of indicating good
convergence properties for LS algorithms. I'm not sure who all has done such
work; the I recall the most about was by Jose Garcia-Luna at SRI.

	Noel